main.py

import numpy as np
import cv2 as cv
import math

#   the MPI model(16 parts):
#   http://posefs1.perception.cs.cmu.edu/OpenPose/models/pose/mpi/pose_iter_160000.caffemodel
#   https://raw.githubusercontent.com/opencv/opencv_extra/master/testdata/dnn/openpose_pose_mpi_faster_4_stages.prototxt

#   (to simplify this sample, the body models are restricted to a single person.)

#   the hand pose model:
#   http://posefs1.perception.cs.cmu.edu/OpenPose/models/hand/pose_iter_102000.caffemodel
#   https://raw.githubusercontent.com/CMU-Perceptual-Computing-Lab/openpose/master/models/hand/pose_deploy.prototxt

threshold = 0.1

HAND_PARTS = {
    "BOTTOM": 0, "BOTTOM1": 1,
    "THUMBBOTTOM": 2, "THUMBMID": 3, "THUMBTOP": 4,
    "INDEXBOTTOM": 5, "INDEXMID1": 6, "INDEXMID2": 7, "INDEXTOP": 8,
    "MIDBOTTOM": 9, "MIDMID1": 10, "MIDMID2": 11, "MIDTOP": 12,
    "RINGBOTTOM": 13, "RINGMID1": 14, "RINGMID2": 15, "RINGTOP": 16,
    "PINKYBOTTOM": 17, "PINKYMID1": 18, "PINKYMID2": 19, "PINKYTOP": 20, "HEAVEN": 21
}

POSE_PAIRS = [
    ["BOTTOM", "BOTTOM1"],     ["BOTTOM1",     "THUMBBOTTOM"], ["THUMBBOTTOM", "THUMBMID"],  ["THUMBMID",  "THUMBTOP"],
    ["BOTTOM", "INDEXBOTTOM"], ["INDEXBOTTOM", "INDEXMID1"],   ["INDEXMID1",   "INDEXMID2"], ["INDEXMID2", "INDEXTOP"],
    ["BOTTOM", "MIDBOTTOM"],   ["MIDBOTTOM",   "MIDMID1"],     ["MIDMID1",     "MIDMID2"],   ["MIDMID2",   "MIDTOP"],
    ["BOTTOM", "RINGBOTTOM"],  ["RINGBOTTOM",  "RINGMID1"],    ["RINGMID1",    "RINGMID2"],  ["RINGMID2",  "RINGTOP"],
    ["BOTTOM", "PINKYBOTTOM"], ["PINKYBOTTOM", "PINKYMID1"],   ["PINKYMID1",   "PINKYMID2"], ["PINKYMID2", "PINKYTOP"]
]

protofile = "./proto/hand_pose.prototxt"
modelfile = "./models/pose_iter_102000.caffemodel"

inWidth = 368
inHeight = 368
net = cv.dnn.readNetFromCaffe(protofile, modelfile)

cap = cv.VideoCapture(0)
while True:
    ret, frame = cap.read()
    frame = cv.flip(frame, 1)

    frameWidth = frame.shape[1]
    frameHeight = frame.shape[0]
    inp = cv.dnn.blobFromImage(frame, 1.0 / 255, (inWidth, inHeight),
                              (0, 0, 0), swapRB=False, crop=False)

    net.setInput(inp)
    out = net.forward()

    if (len(HAND_PARTS) != out.shape[1]):
        continue
    assert(len(HAND_PARTS) == out.shape[1])

    points = []
    for i in range(len(HAND_PARTS)):
        # Slice heatmap of corresponging body's part.
        heatMap = out[0, i, :, :]

        # Originally, we try to find all the local maximums. To simplify a sample
        # we just find a global one. However only a single pose at the same time
        # could be detected this way.
        _, conf, _, point = cv.minMaxLoc(heatMap)
        x = (frameWidth * point[0]) / out.shape[3]
        y = (frameHeight * point[1]) / out.shape[2]

        # Add a point if it's confidence is higher than threshold.
        points.append((int(x), int(y)) if conf > threshold else None)

    for pair in POSE_PAIRS:
        partFrom = pair[0]
        partTo = pair[1]

        assert(partFrom in HAND_PARTS)
        assert(partTo in HAND_PARTS)

        idFrom = HAND_PARTS[partFrom]
        idTo = HAND_PARTS[partTo]

        if points[idFrom] and points[idTo]:
            cv.line(frame, points[idFrom], points[idTo], (0, 255, 0), 3)
            cv.ellipse(frame, points[idFrom], (3, 3), 0, 0, 360, (0, 0, 255), cv.FILLED)
            cv.ellipse(frame, points[idTo], (3, 3), 0, 0, 360, (0, 0, 255), cv.FILLED)

    t, _ = net.getPerfProfile()
    freq = cv.getTickFrequency() / 1000
    cv.putText(frame, '%.2fms' % (t / freq), (10, 20), cv.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 0))

    cv.imshow('OpenPose using OpenCV', frame)

    k = cv.waitKey(30) & 0xff
    if k == 27:
        break